|
|
| |||||||||||||||||||||||||||||||
A cost-effectiveness analysis is usually performed by developing a model of the outcomes of alternative treatments, selecting published data on the probabilities of the outcomes to enter in the model, identifying the expenses associated with each therapy, and then comparing the results (often given as dollars spent per year of life saved) with those of various benchmark therapies. Such analyses vary in their complexity; the data can be derived from a single randomized, controlled clinical trial or from many sources, and the costs incorporated vary extensively from one analysis to another. Most analyses consider the costs of an intervention -- be it a drug, device, procedure, or program -- as well as the costs arising from the disease or condition that may be caused or averted. The costs of hospital stays, laboratory tests, and physicians' follow-up visits are usually included. Some analyses also include the costs of diseases that may ultimately affect those whose lives are prolonged, as well as the wages that may be gained or lost. The approaches to formulating the models and identifying costs vary substantially.
Because of the discretionary nature of the methods used to analyze cost effectiveness and the increasing importance of such analyses (in some countries they are now required before drugs are approved for use), it is incumbent on authors, journal editors, and the funders of these studies to minimize any source of bias. Doing so is particularly important given the tangled financial arrangements that may exist between the authors and the for-profit companies that fund them. For example, we recently saw an analysis for which an author received money directly from the company, another in which the author had a patent pending on an application described in the report, and another in which an author was a consultant to the company that markets the drug that was studied. Although we had no special reason to believe that any of these cost-effectiveness analyses was biased by these financial arrangements, we could not comfortably assume that they were not -- either intentionally or unintentionally. Certainly the financial arrangements created an incentive for bias1,2.
To eliminate any ambiguity, we describe here our recently formulated policy on the publication of cost-effectiveness analyses. As background, we review our long-standing policies on authors' financial conflicts of interest in other types of articles3,4,5. As we explained recently,5 we ask authors of scientific articles who have financial connections with a company that makes the product under study (or its competitors) to inform us of these connections when they submit their manuscripts. We do not make such financial relationships known to reviewers, but we do disclose them to our readers, when appropriate, at the time of publication5. In contrast, we do not even consider review articles or editorials by authors with any financial connections to companies whose products are featured prominently in the article (or their competitors)5. The rationale for these policies is explained elsewhere3,4,5.
In our view, formal cost-effectiveness analyses have some of the features of both original scientific articles and review articles. They are similar to original articles in that the methods and data are explicit and the conclusions are based on the data presented. Yet, they are like review articles in that the assumptions made in constructing the models and the data used in the analysis are usually chosen selectively from the literature, and the choices could be biased. Because of these characteristics, we will treat formal cost-effectiveness analyses partly as we do original articles and partly as we do review articles. Like original articles they will not be excluded from consideration if they are supported by a grant from industry to a nonprofit institution, but like review articles they will be excluded from consideration if any of their authors has a personal financial conflict of interest. Simply disclosing such conflicts of interest, as others have suggested authors do,1 will not suffice.
We will consider for publication any high-quality cost-effectiveness analysis, whether or not the analysis is supported by industry and regardless of whether it uses primary data from a single clinical trial or secondary data from a variety of published sources. When a cost-effectiveness analysis is submitted for publication, we will expect its authors to provide information that will allow us to judge whether an incentive for bias exists. The following conditions must be met: First, any study supported by industry must be funded by a grant to a not-for-profit entity such as a hospital or a university, not to an individual or group of individuals. We will not review such studies if any of the authors is receiving a direct salary from the sponsoring company or a competing company, or if any author has an equity interest in, an ongoing consultancy with, or membership on the scientific advisory board of such a company, or a related patent pending. Second, we must receive written assurance that the agreement between the authors and the funding company ensures the authors' independence in the design of the study, the interpretation of data and writing of the report, and decisions regarding publication, regardless of the results of the analysis. The investigators must retain access to the data. Third, to ensure that the analysis can be assessed or replicated by reviewers and readers, the manuscript must include all the data used in the analysis, all assumptions on which the data are based, and any model used in the analysis. There must be a clear explanation of the assumptions made in building the model. The model must be sufficiently straightforward and lucid so that ordinary readers can comprehend it.
Some will argue that these policies are unnecessarily restrictive. In fact, they differ little from our policies on review articles and editorials, except that they do not prohibit the publication of studies supported by industry. We recognize that bias can compromise even original scientific studies, but we believe that the opportunities for introducing bias into economic studies are far greater, given the discretionary nature of model building and data selection in these analyses. In addition, unlike the effectiveness side of the equation, which is based on biologic phenomena, the cost side is highly artificial. Drug costs, in particular, can be quite arbitrary, since they are prices (not costs) set by the company and are based on criteria known only to the company.
To reduce the likelihood of bias in these analyses, it would be highly desirable to create some distance between the investigators and industry support. If neither the insurance industry nor federal agencies (such as the Food and Drug Administration or the Agency for Health Policy and Research) are willing or able to fund such efforts, an independent entity funded by a consortium of companies in the drug-and-device industry could be created expressly to support economic analyses. Grants from such a fund could support costeffectiveness analyses on a competitive basis and eliminate the concern that support from a single company might lead to hidden biases.
The Journal does not have an extensive history of publishing cost-effectiveness studies, and we certainly do not see ourselves as a repository for many of them. When evaluating such studies for publication, we will also apply our usual criteria: How valid and important are the observations? How interesting will they be to our readers? Lest authors think that any mention of the costs of a treatment or intervention must be confined to formal cost-effectiveness analysis, we also wish to make it clear that these policies do not exclude less formal information about costs in original articles. Such information may be requested by the editors from time to time.
Conflicts of interest in clinical research grow more numerous and problematic as the links between academia and industry grow closer and more complex. The likelihood of bias in cost-effectiveness studies can be reduced if these relations are kept at arm's length.
Jerome P. Kassirer, M.D.
Marcia Angell, M.D.
References
| |||||||||||||||||||||||||||||||
Related Letters:
Cost-Effectiveness Analyses
Steinberg E. P., Powe N. R., Schulman K., Simonich W. L., Tilson H., Kotsanos J. G., Flanagin A., Rennie D., Kassirer J. P., Angell M.
Extract |
Full Text
N Engl J Med 1995;
332:123-125, Jan 12, 1995.
Correspondence
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Previous
